KVM: PPC: Add helpers to call FPU instructions

To emulate paired single instructions, we need to be able to call FPU
operations from within the kernel. Since we don't want gcc to spill
arbitrary FPU code everywhere, we tell it to use a soft fpu.

Since we know we can really call the FPU in safe areas, let's also add
some calls that we can later use to actually execute real world FPU
operations on the host's FPU.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>

1 files changed, 273 insertions, 0 deletions

diff --git a/arch/powerpc/kvm/fpu.S b/arch/powerpc/kvm/fpu.S
new file mode 100644
index 0000000..2b340a3
--- /dev/null
+++ b/arch/powerpc/kvm/fpu.S
@@ -0,0 +1,273 @@
+/*
+ *  FPU helper code to use FPU operations from inside the kernel
+ *
+ *    Copyright (C) 2010 Alexander Graf (agraf@suse.de)
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ *
+ */
+
+#include <asm/reg.h>
+#include <asm/page.h>
+#include <asm/mmu.h>
+#include <asm/pgtable.h>
+#include <asm/cputable.h>
+#include <asm/cache.h>
+#include <asm/thread_info.h>
+#include <asm/ppc_asm.h>
+#include <asm/asm-offsets.h>
+
+/* Instructions operating on single parameters */
+
+/*
+ * Single operation with one input operand
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (short*)&result
+ * R5 = (short*)&param1
+ */
+#define FPS_ONE_IN(name) 					\
+_GLOBAL(fps_ ## name);							\
+	lfd	0,0(r3);		/* load up fpscr value */	\
+	MTFSF_L(0);							\
+	lfs	0,0(r5);						\
+									\
+	name	0,0;							\
+									\
+	stfs	0,0(r4);						\
+	mffs	0;							\
+	stfd	0,0(r3);	/* save new fpscr value */	\
+	blr
+
+/*
+ * Single operation with two input operands
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (short*)&result
+ * R5 = (short*)&param1
+ * R6 = (short*)&param2
+ */
+#define FPS_TWO_IN(name) 					\
+_GLOBAL(fps_ ## name);							\
+	lfd	0,0(r3);		/* load up fpscr value */	\
+	MTFSF_L(0);							\
+	lfs	0,0(r5);						\
+	lfs	1,0(r6);						\
+									\
+	name	0,0,1;							\
+									\
+	stfs	0,0(r4);						\
+	mffs	0;							\
+	stfd	0,0(r3);		/* save new fpscr value */	\
+	blr
+
+/*
+ * Single operation with three input operands
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (short*)&result
+ * R5 = (short*)&param1
+ * R6 = (short*)&param2
+ * R7 = (short*)&param3
+ */
+#define FPS_THREE_IN(name) 					\
+_GLOBAL(fps_ ## name);							\
+	lfd	0,0(r3);		/* load up fpscr value */	\
+	MTFSF_L(0);							\
+	lfs	0,0(r5);						\
+	lfs	1,0(r6);						\
+	lfs	2,0(r7);						\
+									\
+	name	0,0,1,2;						\
+									\
+	stfs	0,0(r4);						\
+	mffs	0;							\
+	stfd	0,0(r3);		/* save new fpscr value */	\
+	blr
+
+FPS_ONE_IN(fres)
+FPS_ONE_IN(frsqrte)
+FPS_ONE_IN(fsqrts)
+FPS_TWO_IN(fadds)
+FPS_TWO_IN(fdivs)
+FPS_TWO_IN(fmuls)
+FPS_TWO_IN(fsubs)
+FPS_THREE_IN(fmadds)
+FPS_THREE_IN(fmsubs)
+FPS_THREE_IN(fnmadds)
+FPS_THREE_IN(fnmsubs)
+FPS_THREE_IN(fsel)
+
+
+/* Instructions operating on double parameters */
+
+/*
+ * Beginning of double instruction processing
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (u32*)&cr
+ * R5 = (double*)&result
+ * R6 = (double*)&param1
+ * R7 = (double*)&param2 [load_two]
+ * R8 = (double*)&param3 [load_three]
+ * LR = instruction call function
+ */
+fpd_load_three:
+	lfd	2,0(r8)			/* load param3 */
+fpd_load_two:
+	lfd	1,0(r7)			/* load param2 */
+fpd_load_one:
+	lfd	0,0(r6)			/* load param1 */
+fpd_load_none:
+	lfd	3,0(r3)			/* load up fpscr value */
+	MTFSF_L(3)
+	lwz	r6, 0(r4)		/* load cr */
+	mtcr	r6
+	blr
+
+/*
+ * End of double instruction processing
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (u32*)&cr
+ * R5 = (double*)&result
+ * LR = caller of instruction call function
+ */
+fpd_return:
+	mfcr	r6
+	stfd	0,0(r5)			/* save result */
+	mffs	0
+	stfd	0,0(r3)			/* save new fpscr value */
+	stw	r6,0(r4)		/* save new cr value */
+	blr
+
+/*
+ * Double operation with no input operand
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (u32*)&cr
+ * R5 = (double*)&result
+ */
+#define FPD_NONE_IN(name) 						\
+_GLOBAL(fpd_ ## name);							\
+	mflr	r12;							\
+	bl	fpd_load_none;						\
+	mtlr	r12;							\
+									\
+	name.	0;			/* call instruction */		\
+	b	fpd_return
+
+/*
+ * Double operation with one input operand
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (u32*)&cr
+ * R5 = (double*)&result
+ * R6 = (double*)&param1
+ */
+#define FPD_ONE_IN(name) 						\
+_GLOBAL(fpd_ ## name);							\
+	mflr	r12;							\
+	bl	fpd_load_one;						\
+	mtlr	r12;							\
+									\
+	name.	0,0;			/* call instruction */		\
+	b	fpd_return
+
+/*
+ * Double operation with two input operands
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (u32*)&cr
+ * R5 = (double*)&result
+ * R6 = (double*)&param1
+ * R7 = (double*)&param2
+ * R8 = (double*)&param3
+ */
+#define FPD_TWO_IN(name) 						\
+_GLOBAL(fpd_ ## name);							\
+	mflr	r12;							\
+	bl	fpd_load_two;						\
+	mtlr	r12;							\
+									\
+	name.	0,0,1;			/* call instruction */		\
+	b	fpd_return
+
+/*
+ * CR Double operation with two input operands
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (u32*)&cr
+ * R5 = (double*)&param1
+ * R6 = (double*)&param2
+ * R7 = (double*)&param3
+ */
+#define FPD_TWO_IN_CR(name)						\
+_GLOBAL(fpd_ ## name);							\
+	lfd	1,0(r6);		/* load param2 */		\
+	lfd	0,0(r5);		/* load param1 */		\
+	lfd	3,0(r3);		/* load up fpscr value */	\
+	MTFSF_L(3);							\
+	lwz	r6, 0(r4);		/* load cr */			\
+	mtcr	r6;							\
+									\
+	name	0,0,1;			/* call instruction */		\
+	mfcr	r6;							\
+	mffs	0;							\
+	stfd	0,0(r3);		/* save new fpscr value */	\
+	stw	r6,0(r4);		/* save new cr value */		\
+	blr
+
+/*
+ * Double operation with three input operands
+ *
+ * R3 = (double*)&fpscr
+ * R4 = (u32*)&cr
+ * R5 = (double*)&result
+ * R6 = (double*)&param1
+ * R7 = (double*)&param2
+ * R8 = (double*)&param3
+ */
+#define FPD_THREE_IN(name) 						\
+_GLOBAL(fpd_ ## name);							\
+	mflr	r12;							\
+	bl	fpd_load_three;						\
+	mtlr	r12;							\
+									\
+	name.	0,0,1,2;		/* call instruction */		\
+	b	fpd_return
+
+FPD_ONE_IN(fsqrts)
+FPD_ONE_IN(frsqrtes)
+FPD_ONE_IN(fres)
+FPD_ONE_IN(frsp)
+FPD_ONE_IN(fctiw)
+FPD_ONE_IN(fctiwz)
+FPD_ONE_IN(fsqrt)
+FPD_ONE_IN(fre)
+FPD_ONE_IN(frsqrte)
+FPD_ONE_IN(fneg)
+FPD_ONE_IN(fabs)
+FPD_TWO_IN(fadds)
+FPD_TWO_IN(fsubs)
+FPD_TWO_IN(fdivs)
+FPD_TWO_IN(fmuls)
+FPD_TWO_IN_CR(fcmpu)
+FPD_TWO_IN(fcpsgn)
+FPD_TWO_IN(fdiv)
+FPD_TWO_IN(fadd)
+FPD_TWO_IN(fmul)
+FPD_TWO_IN_CR(fcmpo)
+FPD_TWO_IN(fsub)
+FPD_THREE_IN(fmsubs)
+FPD_THREE_IN(fmadds)
+FPD_THREE_IN(fnmsubs)
+FPD_THREE_IN(fnmadds)
+FPD_THREE_IN(fsel)
+FPD_THREE_IN(fmsub)
+FPD_THREE_IN(fmadd)
+FPD_THREE_IN(fnmsub)
+FPD_THREE_IN(fnmadd)